Molecular Plant Breeding 2024, Vol.15, No.3, 132-143 http://genbreedpublisher.com/index.php/mpb 136 Figure 2 Box plot of the scores observed over all years and sites for each of the 16 differential hosts (Adopted from Patocchi et al., 2020) Image caption: Severity class 1: No visible lesions; class 2: one or very few lesions detected upon close scrutiny of the tree (±1% of leaves affected); class 3: immediately apparent lesions in general clustered in a few parts of the tree (1 to 5% of leaves affected); class 4: intermediate; class 5: numerous lesions widespread over a large part of the tree (±25%); class 6: intermediate; class 7: severe infection with half of the leaves badly infected by multiple lesions (±50%); class 8: intermediate (±75%); class 9: tree completely affected with (nearly) all the leaves badly infected by multiple lesions (>90%). The bold horizontal line represents the median value. The boxes cover 50% of the middle values while the whiskers each cover 25% of the more divergent values. The dots (outliers) are defined as values located at a distance from the median that is greater than two times the standard deviation (Adopted from Patocchi et al., 2020) 3.3.2 Case 2: accelerating eucalyptus breeding Simiqueli et al. (2023) investigated the practical effectiveness of cross-generational genomic selection (GS) in a recurrent reciprocal selection (RRS) breeding program for hybrid Eucalyptus. The research team employed GBLUP and HBLUP models, utilizing genomic data from approximately 16 000 SNPs to predict the growth volume of G1 and G2 hybrid Eucalyptus seedlings. The results showed that the realized predictive ability (RPA) could exceed 0.80 as the genetic relationship between G1 and G2 increased. Additionally, when the training set included direct parents, the RPA of the GBLUP model exceeded 0.70. This study validated the potential of genomic selection in Eucalyptus breeding, particularly in enhancing genetic gain by shortening breeding cycles and increasing selection intensity, thereby optimizing genotype costs. These findings have significant implications for other plant breeding programs, especially in addressing climate change and promoting sustainable forestry management (Figure 3). Figure 3 illustrates the realized predictive ability (RPA) of different genomic selection models in predicting the mean annual increment (MAI) of G2 hybrid Eucalyptus seedlings. The figure compares three models: GBLUP_G, GBLUP_G+D, and HBLUP, each trained with different datasets. The results indicate that when G1 parents (PARENTS), which are genetically closer to the G2 generation, are used as the training set, all three models show generally higher RPA. Notably, the GBLUP_G model performs best, with RPA values approaching or exceeding 0.80, demonstrating high predictive accuracy. Additionally, the choice of the training dataset significantly impacts predictive ability, underscoring the importance of selecting training data that are closely related to the individuals being predicted to enhance model accuracy.
RkJQdWJsaXNoZXIy MjQ4ODYzMg==